JP2022552498A - Leak detection method - Google Patents

Abstract

The present invention relates to a method for detecting leaks in objects through which a medium flows, in particular pipes or pipelines. According to the present invention, patterns are identified in the observed values of media flow and pressure changes, and temperature changes, and based on the identified patterns, the probability of the presence of a leak due to a self-learning system is determined. be. [Selection drawing] Fig. 1

Description

The invention relates to a leak detection method according to the preamble of claim 1 and to a computer program product according to the preamble of claim 12 .

When a leak occurs in a pipeline, it is often of great economic importance to quickly and safely detect, find, and seal the leak. In particular, for pipeline systems that are typically divided into pipeline segments but have portions that extend between continents and transport large volumes of potentially environmentally harmful products (e.g. crude oil), generally It is also extremely important from an environmental point of view to quickly seal any leaks that do occur.

Based on the principle of conservation of mass, known methods of leak detection generally involve basically forming a mass flow balance. In principle, the amount of transport fluid entering a pipe section should also emerge completely from the end of said pipe section, assuming no leakage in the relevant section. Therefore, under ideal conditions, an imbalance between the incoming mass flow and the outgoing mass flow indicates that there is a leak.

However, this method may not easily ascertain the exact location of the leak along the relevant pipe section. Moreover, this idealized principle may not apply well to real pipeline systems. In particular, the influence of various environmental factors poses a problem. As a result of the considerable length of pipelines or pipeline segments, possibly thousands of kilometers, sections of eg large pipelines often pass through multiple climatic zones. In particular, temperature along the pipeline, which varies from region to region and changes over time, can be a significant amount of interference for ascertaining mass flow balance, depending on the product being transported.

As a result of the thermal expansion of the transported fluid, which in each case depends on the individual product, the mass flow balance can become negative or positive even in the absence of leakage. In this case, the natural volume of the pipeline provides a certain dampening effect. If the pipeline passes through cold regions, the fluid being transported will contract in these areas. Short-term local variations in the ambient temperature of the pipeline, for example, as a result of rainfall or, in the case of agricultural land use on the pipeline, as a result of varying levels of sunlight shading of the ground by tillage and harvesting land. A similar effect is produced by If this effect is ignored when checking mass flow balance, a loss of transported fluid will be recorded despite no leakage.

From an economic point of view, for example, it is almost impossible to completely monitor pipeline networks with large total line lengths and complex branch structures for all relevant factors. Therefore, a leak that does occur is rarely detected immediately by the sensor. Furthermore, environmental factors and thermodynamic properties of the medium are usually not detectable to an adequate degree to allow an accurate statement to be made regarding the correction of the mass flow balance identified for the measurement section. For this reason, various techniques are known for accommodating or compensating for variations caused by statistical processing of ascertained data. This is intended to improve the identification of leaks under real conditions.

To account for thermodynamic changes along pipelines or transported fluids, for example, techniques are needed to model the occurring processes and associated influencing factors with real-time models. A corresponding method is known, for example, under the name "real-time transient model" (RTTM). In some cases, known methods also make it possible to locate leaks in specific areas, for example by detecting propagating pressure waves that appear when a leak occurs.

However, in this case there is always the drawback that the results confirmed by statistical means are often unreliable. This is especially the case when only relatively small amounts of data are available or when a single measurement needs to be evaluated. In the case of an erroneously undetected leak, the high economic and environmental as well as safety risks mean that in cases of suspicion a decision is usually made to perform a manual inspection of the relevant pipeline section. This often requires teams of service engineers to venture long distances into difficult terrain, for example to inspect onshore pipelines. Of course, it is desirable to avoid the associated risks to humans and the environment, as well as potentially significant costs.

Against this background, it is an object of the present invention to improve the reliability of leak detection based on verified data.

The above objects are achieved by a method according to claim 1 and a computer program product according to claim 12. Advantageous developments are in each case the subject matter of the dependent claims.

The proposed method first involves ascertaining a set of values that form the basis for the subsequent evaluation. This value includes at least the flow rate and pressure changes of the product or medium transported by the flow transport object and the temperature value change. Structurally, a flow-carrying object (particularly a pipe or pipeline) is divided into one or more measurement sections. The method includes defining a plurality of measurement points for each measurement section. Preferably, one measuring point each is arranged in the initial area and the final area of the measuring section. Here, the values of the physical quantities mentioned above and, if necessary, the values of further physical quantities are ascertained at each measuring point.

Desired values can be ascertained by direct and/or indirect measurements of physical quantities. In this case, it is advantageous to make the resolution of the recording, in particular the time resolution, as high as possible. Alternatively or additionally, however, data generated in another way, in particular simulated data, can also be applied as values for the method according to the invention or assigned to the measuring points.

It is self-evident that, according to the invention, as an alternative or in addition to ascertaining the absolute values of the relevant physical quantities, it is also possible in each case to ascertain relative values and changes in the applicable physical quantities. Changes, especially changes over time, provide information about the dynamics of the process taking place and are therefore more important for the method than just the absolute value of the physical quantity.

Preferably, the evaluation of whether there is an undesirable loss of media volume is not inherently based on static considerations of actual conditions. Instead, the method according to the invention specifically includes the use of dynamic models. Therefore, primarily changes in the identified quantities, especially changes over time, are of great importance.

The change in the flow rate of the medium, ie the flow rate of the fluid transported in the flow-transporting body, is particularly mass-based, but can also be understood as volume-based. Additionally, pressure changes can relate to hydrostatic pressure and/or dynamic pressure. The underlying temperature value or its change relates in particular to the ambient temperature at the measuring point, but alternatively or additionally can also directly reflect the temperature change of the medium at the measuring point. The recording of pressure and temperature changes is relatively very important, since the medium flow generally varies greatly depending on the existing temperature and/or pressure gradients.

Besides the physical quantities cited, it is also possible to ascertain the values of further quantities, for example the flow rate of the medium or its density, or the external ambient pressure in the area of the measuring point.

If actual measurements are not available or too few quantities are available, values can be interpolated for the desired quantity based on measurements from adjacent or nearby measurement points.

As an alternative or in addition to actual measurements, the values at the measurement points can also be ascertained, in particular by modeling. In particular, flow-transporting objects are modeled in this case, preferably including the flowing medium. This means, for example, that it is possible to simulate the occurrence of certain values of physical quantities of interest and their influence on the flow-transporting object and/or medium can be ascertained on the basis of the underlying model.

Typically, both measured values and values confirmed by modeling or simulation can be subject to fundamental uncertainties, namely random and/or systematic errors. For this reason, the method according to the invention can be used to express in particular in the form of probabilities.

Values can also be generated by forward modeling instead of, or in addition to, real-time models. This includes in particular that an iterative method is adopted, whereby the values available at the measurement points and the influence of such values are predicted. The number of iteration steps can basically be chosen according to what demands are made on the accuracy of the calculation in the individual case.

In the preferred arrangement of modeling employed, possible trends for the system as a whole and/or for individual parameters can be approximated by inter alia ascertaining conditional probability values. In this context, estimation methods such as Bayesian statistical methods and/or maximum likelihood methods, among others, may be included.

In particular, it is possible to model ascertained, modeled and/or simulated data in a one-dimensional model of a flow transport object, preferably a pipeline or pipeline system, in a simple manner. . This may involve selective reduction of data and/or targeted combination of data, in particular to a certain range. This can further be based on weightings to define the extent to which the data used are taken into the model or influence the modeled results. In general, a corresponding simplification to a one-dimensional model allows a much simplified and thus more reliable detection of the important effects that need to be observed.

It is self-evident that high-dimensional models and/or combinations of multiple one-dimensional and/or high-dimensional models can also be employed in an equivalent manner. In principle, the reproduction or use of usually extensive available data for highly simplified models is advantageous for the method according to the invention. A reduced representation according to the current state or likely future trends therein allows, in particular for the user, very reliable detection or evaluation of anomalies in the state and/or operation of the pipeline system. In this case, what level of simplification is finally required can depend, inter alia, on the particular application situation in each individual case.

A real-time model and/or forward modeling of the state of the considered flow transport object is preferably used to determine optimum values for the spatial and/or temporal density of measurement points for capturing the data to be considered. can be used. The measuring point density is in particular unevenly distributed over the entire considered flow transport object or over a particular measuring section. By ascertaining the optimum density, the current and/or future state can be obtained without unnecessary redundant acquisitions resulting in redundant amounts of data whose transmission, storage and processing are time consuming and costly. A sufficient amount of data can be collected locally and/or temporally in relation to the event to assess the . For example, if the special need for reliable assessment of potentially critical situations in high-risk areas is the local provision for higher densities of measurement points, modeling may Each economically optimal extent to which data collection is performed allows to be determined in this regard.

If the values of the underlying physical quantity are ascertained at different measuring points of the measuring section, the method comprises forming at least one numerical value group from these values. Numerical groups can ultimately comprise the total set of recorded or otherwise ascertained values, or can be formed by sub-numeric groups (subgroups) of these.

The ascertained values are supplied to the data processing device for evaluation. The data processing device can be a computer that resides locally in close proximity to the measurement section. However, it is particularly preferred to centrally process data from different measuring points and/or measuring sections by means of a common data processing device. The data processing device may also be a network including multiple interacting computers. In particular, the data processing device is preferably provided at a location physically remote from the monitored measurement section, eg at a central computer center. Data processing can thus also be carried out, for example, on the basis of cloud services principles.

A group of numbers is examined by a data processing device whether the values within the group of numbers form a pattern or form a pattern within the group of numbers. If the data processing device identifies a pattern in the group of values, the pattern, particularly its type and its characteristic strength, can be used to determine the possible presence of a leak in the measured section of the flow-carrying object. This enables a particularly heuristic leak detection, so that leaks occurring in the measurement section under consideration are detected even if evaluation of the available data using known statistical methods does not yield reliable results. can do.

In particular, the method according to the invention is characterized on the one hand by patterns with flow rate changes and/or temperature changes in the medium as a result of environmental influences and on the other hand by leaks or incorrect tapping (Entnahme (Germany), tapping (UK)). , removal) from patterns associated with unwanted loss of flow. The aim here is to be able to react to each situation as quickly as possible in order to keep losses of the transported medium as low as possible.

Preferably, a classification algorithm is applied to a group of numbers or patterns identified in a group of numbers. Thus, existing patterns can not only be identified, but can also be evaluated for classification into different pattern classes. Classes are specifically concerned with the relevance of patterns with respect to the possibility of the presence of leaks.

Alternatively or additionally, pattern analysis algorithms may also be applied to the patterns, which algorithms, in a manner similar to image recognition methods, determine in particular which events are represented by the patterns and how likely they are. Interpret the pattern based on its characteristics to identify

The data processing device is preferably suitably designed to be able to carry out such classification and/or pattern analysis algorithms.

Confirmed values can be stored in a database as a data set. Such data sets can be formed in particular by groups of numerical values that are also used to perform evaluations for pattern identification. Alternatively or additionally, it is preferred that the identified patterns are stored in the database as datasets and/or that such patterns are assigned to previously or in parallel stored datasets. Such patterns and/or underlying values can then be re-accessed for later analysis. In particular, this allows further analysis to be verified.

If a pattern is identified as a result of the evaluation of the values in the formed number group, and if one or more patterns are or have been stored in the database, the patterns can be compared with each other. can. A plurality of stored patterns forms a kind of lookup table, especially in this case. A classification algorithm, optionally applied, can preferably be used to determine which of the stored patterns the newly identified pattern is compared to. If the size of the database of stored patterns, each preferably associated with a particular event, is large enough, the current event can be identified quickly and reliably in this manner according to the principle of fingerprint comparison.

Besides global pattern comparison, it is alternatively or additionally possible to compare merely individual features of a particular pattern defined as characteristic with newly identified patterns. In this case, the characteristic pattern is used as a criterion for the presence of leaks in the measuring section of the flow-carrying object. Characteristic patterns can include, among other things, averaged measurements related to the presence of leaks. Furthermore, it is also possible to use generated data, ie computationally modeled and/or simulated data, to generate the characteristic pattern. In this case, the characteristic pattern preferably corresponds to the pattern that ideally appears in the checked values in the presence of leakage. Depending on the degree of match between the newly identified pattern and the characteristic pattern, the data processing device can be used to make an assertion about the possible presence of leakage in the relevant measurement section. In this case, when a defined threshold is exceeded, this is used inter alia as a hard criterion for the presence of a leak, allowing appropriate action to be initiated, for example a manual check or an emergency shutdown.

A particularly preferred configuration of the method according to the invention provides a data processing device which is used to apply a learning algorithm to the ascertained values or numerical values formed therefrom. Algorithms with the ability to learn not only make the method more useful for current applications, possibly with each iteration, as is the case with commonly used statistical techniques. Rather, the learning algorithm is trained by any application and any treatment of new data. Evolutionary effects make this type of self-learning system more reliable over time. This reduces the error rate for identifying, especially interpreting, patterns of observed values.

Common statistical methods for data analysis related to leak detection are usually directed at compensating for variations that occur so that the desired information can be read from the corresponding adjusted data. . In particular, when algorithms with learning capabilities are used for data analysis, the method according to the invention detects leaks based on the occurrence of appropriate patterns in the observed values, even under conditions where known methods fail. be able to. This may be the case, for example, if the values used have significant outliers, as a result of which the approximations made during statistical processing are far apart. In contrast, the method according to the invention involves the systematic application of empirical data to newly identified values. In particular, the application of algorithms with learning capabilities allows even events not detected by the respective strictly applied statistical algorithm to be identified based on patterns appearing in the values.

In a particularly preferred configuration of the method, the ascertained value or the formed numerical value group is evaluated using an artificial neural network. The data processing device is preferably of suitable design for this purpose.

The learning algorithm is preferably trained using stored values before being applied to the validated value or set of values, said stored values being used to determine the actual presence of an actually occurring event, in particular a leak. or related to events recorded in this connection. Alternatively or additionally, learning algorithms can also be trained based on simulated values. Such simulated values are preferably determined by simulating leakage on flow transfer objects. Training according to the methods described above causes the learning algorithm to learn to associate particular combinations or patterns of numerical values with particular events. Therefore, after appropriate training, using an algorithm with the ability to learn by appropriate construction of queries, it is possible to indicate that there is leakage, especially in the measurement section under consideration, associated with a particular type of event, unknown or It is possible to identify new value patterns.

From a design point of view, it is preferred that the values used in the method according to the invention, in particular the flow rate of the medium, the pressure of the medium and/or the temperature, are ascertained non-invasively in each case. This makes it possible to avoid introducing measuring devices, such as sensors, inside the flow-transporting object and affecting the flow of the medium inside. This creates the risk of distorting the measurement itself and thus the subsequent data evaluation. A suitable measurement of the data is preferably performed by a measuring device placed on or in the shell of the flow-carrying object, eg the wall of a pipeline. In the case of flow, so-called clamp-on flowmeters are particularly suitable, which can detect changes in the flow of the medium inside the flow-transporting object from the outside.

Particularly preferably, the change in flow rate is measured by an acoustic method. This involves ascertaining the flow rate or its change based on the propagation behavior in the flowing medium of an externally introduced acoustic signal. An ultrasound-based method, in which the injected acoustic signal has a moderately high frequency, has been found to be particularly suitable. In particular, the acoustic signal is injected contactlessly, ie without the use of mechanical transducers that externally influence the wall of the flow transport object.

The set of values examined according to the method to identify patterns is formed from the values found at the measurement points, but is not necessarily limited to these values only. It is further possible to include or add to the numerical group further data, in particular commonly available data, for example data relating to the current and/or forecast weather around the flow transport object. This can sometimes make the results of the method according to the invention even more important.

In one preferred configuration of the method, ascertained values from different measurement points are transmitted to a central data processing device. The transmission in this case is preferably done wirelessly.

The present invention further includes a computer program product for determining the likely presence of a leak on a media flow transport object. The computer program product is designed in particular for carrying out the method for leak detection according to the invention or for use in the method according to the invention. Accordingly, the computer program product includes instructions for recognizing patterns in the numerical values, the numerical values being ascertained on the measuring section of the flow-transporting object, and at least the changes in flow rate of the medium, pressure changes in the medium and/or temperature. Formed by the values associated with the change.

The invention is explained in more detail below on the basis of exemplary embodiments. All features described and/or illustrated in the drawings each form an independent aspect of the invention, regardless of their combination in the exemplary embodiment or dependent reference in the claims.

1 shows a schematic representation of an exemplary application situation of the method according to the invention; FIG. Fig. 3 shows a schematic representation of a further application situation of the method according to the invention; 1 is a schematic illustration of the data processing of the method according to the invention; FIG.

FIG. 1 shows a typical application situation for the method according to the invention. In the form of a pipeline or pipeline section for conveying products in the form of a particular fluid medium, the flow-carrying object 1 is laid partly above ground, partly underground and outdoors. .

The illustrated detail represents a measuring section 2 of a fairly long flow-transfer object 1 . The measuring section 2 is monitored by the method for leak detection according to the invention. This is achieved by ascertaining the values of various physical parameters at each of two measurement points 3 .

Starting from the two measuring points 3 shown, the measuring section 2 can also have a larger number of measuring points 3 associated with it. It is certainly more preferred according to the invention, but not essential, that the measuring points 3 for the various physical quantities are arranged at the same position along the measuring section 2 of the flow-transporting object 1 .

Generally, a flow-transporting object 1 can be understood to mean an object through which a medium is basically intended to flow. In this respect, according to the invention it is also possible in principle to ascertain the values for the measuring section 2 where the medium is not continuously flowing. Determining and/or predicting environmental parameters, such as changes in environmental temperature, can be of interest, for example, with respect to upcoming transportation of the medium through the measuring section 2 .

In principle, applicable values for all relevant physical parameters should be determined if possible non-invasively, i.e. without the flowing medium being influenced by components introduced into the flow-transporting object 1 or otherwise. Preferably, the flow is confirmed without being disturbed by the method of .

A value for the change in medium flow rate is ascertained. This is performed in particular by the flow meter 4 . In the embodiment shown in this case a preferred configuration of the flowmeter 4 is shown as a so-called clamp-on flowmeter, which is externally applied to the flow transport object 1 . Therefore, the flow rate of the medium, or changes in that flow rate, can be ascertained non-invasively. It is self-evident that in principle any other type of flow measurement is also useful for confirming the value. Flow rate may be understood as referring to mass and/or volume.

In the illustrated embodiment, the flowmeter 4 is based on an acoustic principle for measuring changes in medium flow. This includes that an acoustic signal, especially in the ultrasonic range, is introduced into the medium through the walls of the flow-transporting object 1 and its propagation velocity is measured in order to draw conclusions about the flow properties of the medium. Preferably, the acoustic signal is a propagating signal that is injected and/or read contactlessly, ie without mechanically coupling the transducer of the flow meter 4 to the wall of the flow-transporting object 1 .

Furthermore, the value of the pressure change in the medium is ascertained at each measuring point 3 . Pressure changes are measured in particular by a suitable pressure sensor 5 .

Furthermore, temperature changes are ascertained, in particular by temperature sensor 6 . This is in particular a value for the ambient temperature or its change at the location of the measuring point 3 . Alternatively or additionally, values can also be recorded at locations remote from the measuring points 3, for example between two measuring points 3 of the measuring section 2. In this connection, such values can be ascertained for the air temperature, the ground temperature, the temperature of the flow-transporting object 1 or the temperature of the flowing medium itself. Thus, in particular the influence of the ambient temperature on the medium in the flow-transporting object 1 along the stretch can be taken into account.

The particularly measured values for the above-mentioned parameters and optionally further physical parameters are then transmitted to the data processing device 7 for further evaluation. The transmission is preferably done wirelessly. It is self-evident that, alternatively or additionally, wire transmission can also take place.

The data processing device 7 can be a central data processing device 7 located remotely from the measurement section 2, as shown in the representation of FIG. Data processing device 7 may be in the form of a single computer, but may also be in the form of a network of multiple interacting computers. Furthermore, it can be provided that the data processing device 7 is configured as a complex system with several computers operating in parallel and/or hierarchically linked.

Data transmission from the measuring point 3 to the data processing device 7 can take place in particular according to common transmission standards such as Bluetooth or WiFi and/or via mobile radio networks. Furthermore, there is also the possibility of satellite-based communication between the measuring points 3 or a device for ascertaining the values provided at the measuring points 3 and the data processing device 7 .

Furthermore, at the measurement point 3, communication can take place between applicable communication devices. By way of example, this makes it possible to provide high performance transmission facilities at only one measuring point 3 or at least at several measuring points 3 for transmitting the ascertained values to the data processing device 7. to The measured values at the individual measuring points 3 of the measuring section 2 are first transmitted over a relatively short distance to such a central measuring point 3 and transferred from there to the data processing device 7 . A suitable design is also by another relay station 10 not associated with a particular measurement point 3, but rather located around the relevant measurement section 2 and thus within range of all communication devices of the relevant measurement point 3. can be realized.

One particular configuration of the method includes at least substantially proprietary data relating to the flow rate of the medium or changes in said flow rate. These data are preferably delivered by the flow meter 4 and/or confirmed by modeling.

Particularly preferably, a network of measuring points 3 or flow meters 4 extending over at least part of the flow-transporting object 1 or over part of the measuring section 2 is also used. In this case, the individual measuring points 3 or flowmeters 4 preferably communicate wirelessly with each other and/or with the data processing device 7 using relay stations 10 arranged between them. Alternatively or additionally, standard mobile radio technology may be relied upon and/or satellite-based communications may be provided, as well as other configurations of the method.

As will be explained in more detail below, the transmitted data are evaluated by the data processing device 7 as part of the method according to the invention for the presence of patterns indicative of the presence of leaks 8 in the examined measuring section 2. be inspected. If such a leak 8 is detected, or if the existence of a leak 8 is sufficiently probable, appropriate action can be taken in a short period of time to provide remedial measures.

In the representation of FIG. 1, such leaks 8 are shown in the portion of the flow transport object 1 that extends underground. The transport medium, which may be crude oil for example, has entered the soil 9 in an uncontrolled manner at the location of the leak 8 and may, for example, contaminate the ground water there. In addition to the economic implications of loss of transport medium, such leaks 8 can have serious ecological consequences. Extensive damage to the environment is not only in the case of catastrophic spills 8, such as large spills of transport medium in a short period of time. Rather, even a small leak 8, which only causes a slow leakage of medium over time, can already pose a significant ecological risk.

FIG. 2 shows, by way of example, a further application of the method according to the invention. The flow-transporting object 1 is formed therein by a relatively complex network of pipes. The details shown are intended to be purely symbolic representations of extensively branched networks of possibly very long pipelines. Apart from branch networks of supply lines, such as over long distances between different regions of the globe, large industrial installations, such as refineries, may also consist of relatively complex pipeline networks. Various measuring sections 2 can be defined in such a highly branched flow-transporting object 1 . The measuring section 2 is not necessarily defined only by the section of the flow-carrying object 1 between two measuring points 3, but can include further areas, in particular provided with three or more measuring points 3. The definition of the measurement section 2 ultimately depends on which measurement points 3 the values received from or ascertained for which measurement points 3 are used for the evaluation by the data processing device 7 .

If the complexity of the branch-like structure of the flow-carrying object 1 is correspondingly high, said object can be difficult to monitor directly by only applicable sensors. As in the case of very long pipelines, complete monitoring of the system ultimately comes at a cost that would be incurred for a reasonable number of sensors. Furthermore, in the various branches of the flow-transporting object 1, the sub-volumes which are in each case fluidly connected to each other provide an interaction and a buffer effect when the medium to be transported propagates through the pipe network. This also prevents the evaluation of mass flow balance.

The method according to the invention has an advantageous effect here by detecting cross-interference events, such as the occurrence of leaks 8 in a particular measurement section 2, by identifying patterns of observed values.

Different temperature influences on the behavior of the transport medium occur not only when passing through different climatic zones or because of different weather conditions along the pipeline. Also in the example of industrial installations, pipelines usually run along structures with different temperatures. For this reason, the temperature of the medium typically changes as it flows through the pipeline or pipeline network. Associated expansion or contraction of the medium can significantly interfere with the determination of the mass flow balance, e.g. Hinder detection of losses.

In this respect, the method according to the invention allows in particular that different influencing factors usually influence the transport medium, in particular the prevailing pressure and/or flow conditions, on different time scales. Changes in climate or weather-related influences generally affect the medium in pipelines, especially underground pipelines, with a time delay, which involves some inertia in the reaction of the system. In contrast, the desired tapping of the medium, for example by the end consumer, results in particularly short-term and particularly locally occurring fluctuations, which likewise need to be taken into account in an appropriate manner.

Particularly desirably, unplanned tapping of the medium in measurement section 2, for example by the final consumer, can be modeled by suitable local consumption measurements and included in the method according to the invention. For this purpose, a suitable positioning of one or more measuring points 3, including in particular the flow meter 4, can be defined in the vicinity of the known tapping point.

The purpose of the method according to the invention is to determine the pattern of ascertained values arising on the basis of temperature and volume fluctuations in the medium due to external and internal influences, the actual flow of the medium from the transport object 1 on the transport path. is to distinguish it from patterns associated with the loss of Natural influences on media are diverse and therefore difficult to fully account for by general statistical methods alone.

The fluctuations that occur are primarily related in particular to changes in the temperature of the medium transported temporally and spatially along the flow-transporting object 1 . It is highly dependent on ambient temperature, but is affected by many other factors. Air and ground temperatures depend on the amount of solar radiation to different extents and affect the temperature of the medium accordingly. Rain and clouds, on the other hand, have a short-term cooling effect. Furthermore, especially in the case of underground pipelines, surface biomass can affect the temperature of the medium in the line, for example in the form of thermal insulation or sunlight being blocked by the ground. Also, this factor is sometimes subject to short-term changes, for example as a result of growing and harvesting in areas used for agriculture.

If the flow-transporting object 1 is of sufficiently large extent or a correspondingly complex branch-like structure, such as a pipeline or pipeline network, thermodynamic changes in the flow properties of the transported medium are also generally internal effects. It always happens for a reason. The reason for this is, for example, fluctuations or changes in the flow resistance due to the geometry of the conduit. Especially when the transport medium is composed of different substances, further variations in its composition may occur. This too can affect the flow behavior of the medium.

Large pipelines or pipe networks are further subjected to so-called "line packing", i.e. filling the conduits with medium to increase the working pressure before the medium exits again at a certain point or is tapped. It can have a significant natural volume that is initially filled. A sufficiently large internal volume of the flow-transporting object 1 also provides a damping effect that allows volume-related changes of the medium to be recorded only indirectly, even during operation. Therefore, without further consideration of internal and/or external parameters, it is almost impossible to draw meaningful conclusions from comparative measurements of the flow rate or its changes at the input and output of the measuring section 2 of the flow-transporting object 1. .

Extensive testing has surprisingly shown that the observed values can form different kinds of patterns. Even with the inclusion of environmental parameters, there are natural variations that cannot be completely removed by common statistical methods, but patterns occur in the data. These are distinguished from patterns that can be observed in the case of real mass losses, for example as a result of leaks 8, line breaks or incorrect tapping of the medium on the transport path.

This is the starting point of the present invention in that these two kinds of patterns are identified and distinguished from each other. As already mentioned, the method uses the data processing device 7 to analyze these during or after the evaluation of the ascertained values for changes in flow rate, pressure and temperature, and optionally further changes in physical quantities. involves looking for patterns in the values of .

The representation shown in FIG. 3 shows the basic sequence for the evaluation of the ascertained values by the data processing device 7 for leak detection. First, a numerical group 11 is formed from the ascertained values and analyzed by the data processing device 7 for the presence of patterns. The numerical value group 11 can consist of all of the values ascertained at the measuring point 3 of the measuring section 2, or can be a subset thereof.

If a pattern is identified in the numerical value group 11 , the data processing device 7 can use this pattern as a basis for determining the possible presence of a leak 8 in the relevant measurement section 2 of the flow-carrying object 1 . Such patterns in the data of numeric group 11 are identified by suitable algorithms of the detection routine, especially as in digital image recognition.

The data processing device 7 is preferably designed to carry out a classification algorithm and applies such algorithm to the number group 11 . Identified patterns are thus classified with respect to their type, nature and/or characteristics.

Alternatively or in addition to such a classification algorithm, a pattern analysis algorithm can also be applied to the numeric group 11 by the data processing device 7 . Such pattern analysis algorithms can interpret the importance of identified patterns. This allows us to make assertions about what real-world events the patterns occurring in the observed values represent.

In one preferred arrangement, the data processing device 7 accesses a database in which identified patterns may be stored as data sets 12 . The same is true for the confirmed values, namely number group 11. In particular, confirmed patterns, numerical values 11, and/or specific truly occurring events, such as the presence of leaks 8, can be linked together and stored in a database as data sets 12 or as joint data sets 12. can.

A comparison of the analyzed numerical value group 11 pattern with one or more patterns stored in the data set 12 of the database allows, in the simplest case, to rapidly assign the identified pattern to the event group. can. Such a comparison by fingerprinting methods, inter alia, allows the data processing device 7 to access a data set 12 classified with respect to stored patterns and/or related events, and identify patterns based on their characteristics. It is possible if it can be uniquely assigned to one of the classes.

Alternatively or additionally, the characteristic or idealized pattern determines the possible presence of a leak 8 in the measurement section 2 under consideration by comparison with the pattern identified in the group of values 11. It can also be used as a reference taken as a basis for Such characteristic patterns may be based on measurements from one or more measurements related to events that actually occur, or may be based on simulated values.

If there is a sufficient degree of agreement between the identified pattern and the characteristic pattern, i.e. above the defined threshold, the criteria for the presence of leak 8 are evaluated as fulfilled so that appropriate action can be taken. can be

A configuration of the method according to the invention in which the data processing device 7 applies a learning algorithm to the ascertained values or numerical values 11 in order to identify patterns is particularly preferred. Alternatively or additionally, algorithms with learning capabilities can be used to function as classification algorithms and/or pattern analysis algorithms. Compared to the above-described method, which inherently identifies and evaluates patterns in number group 11 based on tightly defined criteria, algorithms with learning ability will, as a result of appropriate training with appropriate data, It has the advantage of being more powerful and more reliable. Therefore, the susceptibility to errors in misinterpreting patterns as indicators of leaks 8 (false positives) and not detecting existing leaks 8 based on confirmed values (false positives) is reduced. Become.

Such a learning algorithm is preferably trained with data sets 12 measured when real events occur, in particular events associated with or associated with the presence of leaks 8 . Such data ultimately model reality in the best possible way, so that the trained learning algorithm ultimately yields confirmed values in individual cases under real conditions. Tailored to the specific pattern obtained.

Alternatively or additionally, learning algorithms can be trained using simulated values or model data. This allows adding components to the algorithm that relate to idealized conditions.

For optimal detection performance with respect to identifying, classifying and/or interpreting patterns in numerical values 11, it may be particularly advantageous to train the algorithm on a combination of real and simulated or ideal data. There are cases.

In a more preferred arrangement, the data processing device 7 can use a particularly iterative method for modeling the values. This includes using methods specifically for forward modeling to ascertain values that can be expected under specific operating and/or ambient conditions.

Values determined by way of such modeling methods can be adopted in different ways for the method according to the invention. By way of example, parallel application of such modeling methods may allow for independent verification of the measurements and/or patterns appearing in the measurements.

Data obtained by forward modeling methods are also suitable for training learning algorithms.

Preferably, a comparison of the evaluation of real data and the modeling of certain trends of the system allows possible artifacts of pattern identification to be determined and, in particular, corrected. In this way, it is possible to preferably compensate for the shortcomings of the learning algorithm that appear in this context, in particular subject to non-optimal prioritization during training of the algorithm. Therefore, repeated use of this approach continuously improves the reliability of pattern identification.

Furthermore, it is possible to continuously link the pattern recognition-based method according to the invention with a corresponding method for modeling data based on measurements. This allows, for example, the risk of imminent structural failure of the flow transport object 1 by modeling future trends on the basis of known or measured starting parameters and identifying patterns occurring in the results thus obtained. can be evaluated.

In addition to considering the dataset 12 of the database in the manner described above, it is also possible to include data from external sources, in particular publicly available data, in a different manner. These data are added to and/or linked with the value group 11, in particular to be considered for the evaluation. However, this kind of external data can also be used for modeling and/or training of learning algorithms. By way of example, the data may relate to the weather, the geological composition of the ground, the area's particular agricultural use, and the like.

Evaluating the ascertained values, or the numerical value group 11 formed therefrom, identifies patterns in the numerical value group 11 and, if necessary, interprets the pattern or otherwise identifies it as a specific event or event. This includes associating with likelihood. Preferably, the data processing device 7 then produces a suitable output 13 conveying to the user the results of the preceding analysis or of the method used.

Output 13 may be provided in different ways, preferably in visual, audible and/or textual form. In particular, as shown in FIG. 3, output 13 may consist of a warning regarding the presence of leak 8 . Additionally, for example, a status report may be generated.

For systems that operate automatically, alternatively or additionally, remedial action regarding output 13 may be taken immediately, e.g. a warning delivered to maintenance and/or service personnel.

In this case, it is also self-evident that in principle it is also possible to combine different outputs 13 or reactions for the analytical results of the method.

1 flow transfer object 2 measurement section 3 measurement point 4 flow meter 5 pressure sensor 6 temperature sensor 7 data processing device 8 leak 9 soil 10 relay station 11 number group 12 data set 13 output

Claims (12)

A method for detecting leaks in a medium flow-transporting object (1), in particular a pipe or pipeline, at each of a plurality of measuring points (3) on a measuring section (2) of said flow-transporting object (1). , values relating to changes in the flow rate of said medium, changes in pressure in said medium and changes in temperature values are ascertained, said ascertained values being recorded and statistically evaluated by a data processing device (7) ,
A number group (11) is formed from said ascertained values, a pattern is identified in said number group (11) by said data processing device (7), and based on said identified pattern, said flow transport object (1 ), wherein the possible presence of a leak (8) in the measuring section (2) of ) is determined. Method according to claim 1, characterized in that a classification algorithm and/or a pattern analysis algorithm are applied to the identified patterns. Method according to claim 1 or 2, characterized in that the identified patterns are stored as datasets (12) in a database and/or assigned to datasets (12) stored in the database. . Method according to one of claims 1 to 3, characterized in that the identified pattern is compared with one or more stored patterns. 5. According to one of claims 1 to 4, characterized in that the characteristic pattern serves as a criterion for the presence of leaks (8) in the measuring section (2) of the flow-transporting object (1). the method of. Method according to one of claims 1 to 5, characterized in that said data processing device (7) applies a learning algorithm to said ascertained values and/or said numerical group (11). . Said learning algorithm is trained using stored values and/or simulated values before being applied to said confirmed values and/or said set of values (11), said stored values and/or or the simulated value is related to the actual and/or simulated presence of a leak (8) on the medium flow transport object (1). Method. The method according to any one of claims 1 to 7, characterized in that the values relating to changes in the medium flow rate, the values relating to the pressure changes in the medium and/or the changes in the temperature values are each non-invasively ascertained. 1. The method according to item 1. Method according to one of the preceding claims, characterized in that the change in flow rate is measured by an acoustic, preferably ultrasound-based method. Data from external sources, in particular data relating to the weather around said measuring section (2) and/or measuring points (3), are processed by said data processing device (7) and in particular linked to said group of values (11). , and/or added to the numerical group (11). 11. The method according to one of claims 1 to 10, characterized in that said ascertained values from different measuring points (3) are transmitted to a central data processing device (7), preferably wirelessly. Method. A computer program product for determining the possible presence of a leak (8) on a medium flow transport object (1), in particular a pipe or pipeline, comprising:
comprising instructions for recognizing a pattern in a group of numbers (11), said group of numbers (11) being associated with changes in flow rate of said medium, pressure changes in said medium and/or temperature changes in said flow-transporting object (1); A computer program product characterized in that it is formed by the values ascertained on the measurement section (2).

Images